PROJECT SUMMARY (30 LINES) Tumors are strongly dependent on the aberrant activity of the tumor checkpoint modules that are responsible for maintaining tumor cell-state or for inducing drug resistance. Thus inference of compounds and combinations that can collapse the activity of such tumor checkpoints in vivo is extremely relevant toward developing novel cancer therapeutics and rescuing drugs that induce relapse. The overall goal of this project is explore the hypothesis that compounds or combinations that are computationally inferred to induce checkpoint collapse (in vitro or in vivo) will translate to preclinical models, in terms of inducing tumor regression or sensitivity rescue in vivo. Specifically, the project will aim at developing and experimentally validating novel algorithms to: (1) Elucidate compound specific targets and effectors. This will be done under aim 1 by extending the current VIPER and DeMAND algorithms as well as by developing additional structure- and network-based algorithms to identify specific pathways and proteins whose activity is dysregulated by compound activity. (2) Study synergistic compound activity from RNASeq profiles of individual compound perturbations. This will be done under aim 2 by specifically studying and targeting the distinct mechanisms that underlie synergistic drug activity. Developed algorithms will be used to prioritize combination therapy to treat PDX models from the N of 1 pilot study. Finally developing a predictive framework for the assessment of compound and compound combination toxicity. This will be done under aim 3 by integrating results from the analysis of gene expression profiles following in vivo perturbations with established toxic compounds both from publicly available databases and novel assays.